If you want to see the future of Big Data, look no further than the nearest gaming-development studio. It isn't all fun and first-person-shooting. Game developers are the sentinels of a variety of advanced IT techniques, placing them in front of the general IT population with regard to using real-time analytics and cloud computing, among other areas. Computer games are big business, with last year’s worldwide sales north of $20 billion (and even the subcategory of social games at over $2 billion)—compare that to the $8 billion per year that Americans spend on movie tickets. SlashBI previously looked at how Riot Games is using Hadoop and other NoSQL tools to track its players’ statistics and improve gameplay. It’s not the only game studio taking technology to new heights. Gamers have long been ahead of the curve in three key areas: rapid changes in computing infrastructure, persistent and more personalized data connections from the cloud, and a long history of using graphical processors (GPUs) to support high performance computing. Let's look at each of these items in more detail.
Rapid On-Demand Computing Changes
"From an infrastructure perspective, games have a high volume of data points due to user interactions and typically have a unique need for fast response. This makes them very tricky cloud data users," said Robert Nelson, the CEO of Broken Bulb Studios, which develops mobile and Facebook-based games. The company makes use of SoftLayer for their cloud hosting and stores several terabytes of data; peak transfer rates can reach over 200 Mbps. "SoftLayer's platform has a unique combination of scalability and customizability, which supports the dynamic infrastructure of gaming companies. SoftLayer can provision cloud computing instances in minutes, allowing us to rapidly scale up or down as our needs change," he added. As part of a new game launch last year, the company saw 1.4 million players come to its Website in a week, up from a few thousand beta-users prior to the launch. Heavy demand for some games has driven SoftLayer to double its infrastructure overnight. This variation in demand is the wheelhouse of general cloud computing, but games do seem to have a higher fluctuation than traditional IT applications. When another gaming studio, Hothead Games, launched its Big Win series of sports games last year, its required number of servers on SoftLayer’s hosting network rose from six to 60—and it was handled with ease. "Our code makes hundreds of millions of database transactions a day. It's critical to our business that every single one of those works reliably and is super fast," said Joel DeYoung, director of technology at Hothead Games. SoftLayer isn't alone in recognizing this market. Peer 1 Hosting has also worked with some of the world's largest game developers to seamlessly deliver games despite wide fluctuations in demand. One launch saw traffic spike to more than a thousand servers, which were automatically provisioned by their managed hosting service. There’s also Joyent, which hosts Quizlet, one of the largest e-learning games. Thanks to some careful analysis, Quizlet found too many PHP calls and was able to rewrite code to speed up operations; as a result, it has scaled up from a few hundred beta-users several years ago to more than 60 million page-views a month today.
Persistence and Personalization
"Gaming is a more interesting target market than traditional B2C spaces," said Brian Stone of Causata, a customer experience management company. "And online gaming is even more so, since it offers unparalleled opportunities for cross-selling and upselling. You are competing with your friends and constantly checking your play statistics, and very involved with your social network. Compare that to an online banking app: the game is a lot more engaging and personal." But underlying analysis for this personalization requires solid support for various business-intelligence tools, which many game studios have fully embraced. IsCool Entertainment can analyze the immense dimensions of its online gamers’ activity and social behavior with Actian's Vectorwise, a Hadoop analytics engine. This provides data for calculating rewards, generating leader boards and delivering virtual prizes—all to enhance customer engagement and retention. "Games have a persistent connection with the user and as a result, we get so much more data," said Reid Tatoris, the CEO of PlayThru.com. The company produces games that can be used in place of annoying Captcha Turing tests to verify that an actual human is signing up for a website. "Interacting with an app doesn't give you the how. Is this person's interaction human? How did you go about completing the task? That is where we try to help."
PlayThru has collected lots of insights into personal preferences as a result of deploying their app across 20 million page views. "When you are playing a mobile game,” Tatoris said, “you can get all sorts of information about what the user is doing, where they are located, and how they are interacting with the game in near real-time." Try getting that level of insight from a user creating a Word document. As a result of those games, PlayThru is seeing submission rates increase by 40 percent over the traditional text-based Captcha applications. One game that champions personalization is the site Fanhood.com, which connects sports fans with their favorite teams via Facebook. "There is so much content to navigate, we try to focus on what is relevant for a particular fan," said company CEO Brandon Ramsey. "What's more, we try to structure it within your Facebook social graph so you can immediately tell which of your friends are fans of teams that your local team is playing this week." Fanhood uses MongoDB and Cassandra to manage millions of rows of data for each team and fan to create its personal team updates. Having all this data is a tremendous opportunity if managed properly. Causata handles an online sports betting site, and can provide all sorts of specifics such as who opens which emails and the paths that customers take within the site. It can even predict the number of bets made and their value, the average duration between bets, and the sports most likely to interest certain visitors. Causata builds these models using R and Hadoop.
GPU Computing
Finally, there’s the notion of using graphics processors to boost general computing tasks. The concept isn’t new: several years ago, a group of Swiss researchers put together a cluster of 200 PS3s to form a primitive supercomputer. While Sony disabled this ability soon afterwards, a number of hosting providers now offer on-demand GPU computing in the cloud, making use of Nvidia graphics processors and specialized Linux operating systems that can take advantage of this increased horsepower. Providers include Amazon Web Services, Peer 1 and SoftLayer, among others. (One of the Amazon configurations was able to place in the top500 list of the most powerful supercomputers for several years running.) All of these BI tools and advanced computing techniques have brought about what Kimberly Chulis, the CEO of Core Analytics, called "a new focus on advanced analytics and micro-segmentation to drive player monetization. Game developers and brands have an opportunity to apply these big data analytics techniques to capture rich and varied behavioral and multi-structured game and player data." Images: TellTale Games, PlayThru.com
